Multilayer neural networks and adaptive nonlinear control of agile anti-air missiles

摘要

Research has shown that neural networks can be used to improve upon approximate dynamic inversion for control of uncertain nonlinear systems. ^In one architecture, the neural network adaptively cancels inversion errors through on-line learning. ^Such learning is accomplished by a simple weight update rule derived from Liapunov theory, thus assuring the stability of the closed-loop system. ^We review this methodology and extend it to incorporate an important class of neural networks with a single hidden layer. ^An agile antiaircraft missile autopilot is subsequently designed using this control scheme. ^First, a control law based on approximate inversion of the nonlinear dynamics is presented. ^This control system is augmented by the addition of a multilayer neural network with on-line learning. ^Finally, numerical results from a nonlinear agile antiaircraft missile simulation demonstrate the effectiveness of the resulting autopilot. ^(Author)